Training in progress, step 1000

config.json

@@ -31,7 +31,7 @@
   "mask_time_length": 10,
   "mask_time_min_masks": 2,
   "mask_time_prob": 0.05,
-  "max_length": 448,
+  "max_length": null,
   "max_source_positions": 1500,
   "max_target_positions": 448,
   "median_filter_width": 7,
@@ -41,7 +41,7 @@
   "pad_token_id": 50256,
   "scale_embedding": false,
   "torch_dtype": "float32",
-  "transformers_version": "4.46.0.dev0",
+  "transformers_version": "4.49.0.dev0",
   "use_cache": false,
   "use_weighted_layer_sum": false,
   "vocab_size": 51866

model-00001-of-00002.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:70e95881b347aab16bfadb8b06b24d389bceb51ce8b9894db264ed67fc19f29e
+oid sha256:6f19f041acff32cec525af7f2685700fdfef887382388ed14ec888a4fbf1eafe
 size 4993448880

model-00002-of-00002.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:52bad486fd09fef8dae53b1e1e2b13ea27a7a645786a7a9920fe7344b96bf162
+oid sha256:9e392c9b8208f172dbb0775e074de0beba10a38023b69cfe310e8637407f430c
 size 1180663192

run.sh

@@ -1,11 +1,10 @@
 WANDB_PROJECT=whisper-medium-eu \
 	python run_speech_recognition_seq2seq_streaming.py \
 	--model_name_or_path="openai/whisper-large-v3" \
-	--dataset_name="mozilla-foundation/common_voice_17_0" \
-	--dataset_config_name="eu" \
+	--dataset_name="asierhv/composite_corpus_eu_v2.1" \
 	--language="basque" \
-	--train_split_name="train+validation" \
-	--eval_split_name="test" \
+	--train_split_name="train" \
+	--eval_split_name="dev_parl+test_parl+test_cv+test_oslr" \
 	--model_index_name="Whisper Large Basque" \
 	--max_steps="10000" \
 	--output_dir="./" \
@@ -14,7 +13,7 @@ WANDB_PROJECT=whisper-medium-eu \
 	--gradient_accumulation_steps="1" \
 	--logging_steps="25" \
 	--learning_rate="4.375e-6" \
-	--warmup_steps="500" \
+	--warmup_steps="1000" \
 	--evaluation_strategy="steps" \
 	--eval_steps="500" \
 	--save_strategy="steps" \
@@ -31,7 +30,6 @@ WANDB_PROJECT=whisper-medium-eu \
 	--gradient_checkpointing \
 	--fp16 \
 	--overwrite_output_dir \
-	--resume_from_checkpoint="checkpoint-9000" \
 	--do_train \
 	--do_eval \
 	--predict_with_generate \
@@ -39,4 +37,4 @@ WANDB_PROJECT=whisper-medium-eu \
 	--streaming \
 	--push_to_hub \
 	--report_to "wandb" \
-	--run_name "whisper-large-eu"
+	--run_name "whisper-large-eu-v3"

run_speech_recognition_seq2seq_streaming.py

@@ -25,6 +25,7 @@ import os
 import sys
 from dataclasses import dataclass, field
 from typing import Any, Dict, List, Optional, Union
+import numpy
 
 import datasets
 import torch
@@ -265,25 +266,58 @@ class DataCollatorSpeechSeq2SeqWithPadding:
         return batch
 
 
-def load_maybe_streaming_dataset(dataset_name, dataset_config_name, split="train", streaming=True, **kwargs):
+def load_maybe_streaming_dataset(dataset_name, dataset_config_name, data_args, split="train", streaming=True, **kwargs):
     """
-    Utility function to load a dataset in streaming mode. For datasets with multiple splits,
-    each split is loaded individually and then splits combined by taking alternating examples from
-    each (interleaving).
+    Utility function to load a dataset in streaming mode.
     """
+    logger.info(f"Loading dataset {dataset_name} split {split} (streaming={streaming})")
+    
+    def load_single_split(split_name):
+        logger.info(f"Loading split: {split_name}")
+        ds = load_dataset(
+            dataset_name, 
+            dataset_config_name, 
+            split=split_name, 
+            streaming=streaming, 
+            trust_remote_code=True, 
+            **kwargs
+        )
+        
+        # Add validation transform to ensure consistent audio format
+        def validate_example(example):
+            if not isinstance(example[data_args.audio_column_name], dict):
+                example[data_args.audio_column_name] = {
+                    'array': example[data_args.audio_column_name].array,
+                    'sampling_rate': example[data_args.audio_column_name].sampling_rate,
+                    'path': getattr(example[data_args.audio_column_name], 'path', None)
+                }
+            return example
+            
+        ds = ds.map(validate_example)
+        
+        # Log first example structure for debugging
+        try:
+            first_example = next(iter(ds))
+            logger.info(f"First example from {split_name}:")
+            logger.info(f"  Keys: {first_example.keys()}")
+            if data_args.audio_column_name in first_example:
+                audio = first_example[data_args.audio_column_name]
+                logger.info(f"  Audio type: {type(audio)}")
+                if isinstance(audio, dict):
+                    logger.info(f"  Audio keys: {audio.keys()}")
+                    logger.info(f"  Array type: {type(audio['array']) if 'array' in audio else 'missing'}")
+        except Exception as e:
+            logger.warning(f"Could not inspect first example from {split_name}: {e}")
+            
+        return ds
+    
     if "+" in split:
-        # load multiple splits separated by the `+` symbol with streaming mode
-        dataset_splits = [
-            load_dataset(dataset_name, dataset_config_name, split=split_name, streaming=streaming, trust_remote_code=True, **kwargs)
-            for split_name in split.split("+")
-        ]
-        # interleave multiple splits to form one dataset
-        interleaved_dataset = interleave_datasets(dataset_splits)
-        return interleaved_dataset
+        # Load and validate each split individually
+        dataset_splits = [load_single_split(split_name) for split_name in split.split("+")]
+        # Interleave datasets
+        return interleave_datasets(dataset_splits)
     else:
-        # load a single split *with* streaming mode
-        dataset = load_dataset(dataset_name, dataset_config_name, split=split, streaming=streaming, trust_remote_code=True, **kwargs)
-        return dataset
+        return load_single_split(split)
 
 
 def main():
@@ -356,37 +390,39 @@ def main():
         raw_datasets["train"] = load_maybe_streaming_dataset(
             data_args.dataset_name,
             data_args.dataset_config_name,
+            data_args,
             split=data_args.train_split_name,
-            # xezpeleta
-            #use_auth_token=True if model_args.use_auth_token else None,
             streaming=data_args.streaming,
         )
 
+        # Get features from train dataset since it's guaranteed to exist if do_train is True
+        train_dataset = raw_datasets["train"]
+        first_example = next(iter(train_dataset))
+        raw_datasets_features = list(first_example.keys())
+
+        if data_args.audio_column_name not in raw_datasets_features:
+            raise ValueError(
+                f"--audio_column_name '{data_args.audio_column_name}' not found in dataset '{data_args.dataset_name}'. "
+                "Make sure to set `--audio_column_name` to the correct audio column - one of "
+                f"{', '.join(raw_datasets_features)}."
+            )
+
+        if data_args.text_column_name not in raw_datasets_features:
+            raise ValueError(
+                f"--text_column_name {data_args.text_column_name} not found in dataset '{data_args.dataset_name}'. "
+                "Make sure to set `--text_column_name` to the correct text column - one of "
+                f"{', '.join(raw_datasets_features)}."
+            )
+
     if training_args.do_eval:
         raw_datasets["eval"] = load_maybe_streaming_dataset(
             data_args.dataset_name,
             data_args.dataset_config_name,
+            data_args,
             split=data_args.eval_split_name,
-            #use_auth_token=True if model_args.use_auth_token else None,
             streaming=data_args.streaming,
         )
 
-    raw_datasets_features = list(next(iter(raw_datasets.values())).features.keys())
-
-    if data_args.audio_column_name not in raw_datasets_features:
-        raise ValueError(
-            f"--audio_column_name '{data_args.audio_column_name}' not found in dataset '{data_args.dataset_name}'. "
-            "Make sure to set `--audio_column_name` to the correct audio column - one of "
-            f"{', '.join(raw_datasets_features)}."
-        )
-
-    if data_args.text_column_name not in raw_datasets_features:
-        raise ValueError(
-            f"--text_column_name {data_args.text_column_name} not found in dataset '{data_args.dataset_name}'. "
-            "Make sure to set `--text_column_name` to the correct text column - one of "
-            f"{', '.join(raw_datasets_features)}."
-        )
-
     # 5. Load pretrained model, tokenizer, and feature extractor
     #
     # Distributed training:
@@ -438,14 +474,12 @@ def main():
         tokenizer.set_prefix_tokens(language=data_args.language, task=data_args.task)
 
     # 6. Resample speech dataset if necessary
-    dataset_sampling_rate = next(iter(raw_datasets.values())).features[data_args.audio_column_name].sampling_rate
-    if dataset_sampling_rate != feature_extractor.sampling_rate:
-        raw_datasets = raw_datasets.cast_column(
-            data_args.audio_column_name, datasets.features.Audio(sampling_rate=feature_extractor.sampling_rate)
-        )
+    # For streaming datasets with audio bytes, sampling rate is handled in prepare_dataset
+    logger.info("Using feature extractor sampling rate: %d", feature_extractor.sampling_rate)
+    dataset_sampling_rate = feature_extractor.sampling_rate
 
     # 7. Preprocessing the datasets.
-    # We need to read the audio files as arrays and tokenize the targets.
+    logger.info("Starting dataset preprocessing")
     max_input_length = data_args.max_duration_in_seconds * feature_extractor.sampling_rate
     min_input_length = data_args.min_duration_in_seconds * feature_extractor.sampling_rate
     audio_column_name = data_args.audio_column_name
@@ -469,20 +503,59 @@ def main():
             else raw_datasets["eval"].select(range(data_args.max_eval_samples))
         )
 
+    # Inspect dataset before processing
+    for split, dataset in raw_datasets.items():
+        try:
+            first_example = next(iter(dataset))
+            logger.info(f"First example from {split} before processing:")
+            logger.info(f"Keys: {first_example.keys()}")
+            if audio_column_name in first_example:
+                audio_data = first_example[audio_column_name]
+                logger.info(f"Audio column type: {type(audio_data)}")
+                if isinstance(audio_data, dict):
+                    logger.info(f"Audio keys: {audio_data.keys()}")
+        except Exception as e:
+            logger.warning(f"Could not inspect first example from {split}: {e}")
+
     def prepare_dataset(batch):
-        # process audio
-        sample = batch[audio_column_name]
-        inputs = feature_extractor(sample["array"], sampling_rate=sample["sampling_rate"])
-        # process audio length
-        batch[model_input_name] = inputs.get(model_input_name)[0]
-        batch["input_length"] = len(sample["array"])
-
-        # process targets
-        input_str = batch[text_column_name].lower() if do_lower_case else batch[text_column_name]
-        if do_remove_punctuation:
-            input_str = normalizer(input_str).strip()
-        batch["labels"] = tokenizer(input_str).input_ids
-        return batch
+        try:
+            # Validate audio format
+            audio = batch[audio_column_name]
+            
+            # Load audio from bytes if needed
+            if isinstance(audio, dict) and 'bytes' in audio:
+                import io
+                import soundfile as sf
+                audio_bytes = io.BytesIO(audio['bytes'])
+                audio_array, sampling_rate = sf.read(audio_bytes)
+                audio = {'array': audio_array, 'sampling_rate': sampling_rate}
+            
+            # Process audio through feature extractor
+            inputs = feature_extractor(audio['array'], sampling_rate=audio['sampling_rate'])
+            batch["input_length"] = len(audio['array'])
+            batch[model_input_name] = inputs[model_input_name][0]
+            
+            # Process text
+            input_str = batch[text_column_name].lower() if do_lower_case else batch[text_column_name]
+            if do_remove_punctuation:
+                input_str = normalizer(input_str).strip()
+            batch["labels"] = tokenizer(input_str).input_ids
+            
+            return batch
+            
+        except Exception as e:
+            logger.error(f"Error processing batch in prepare_dataset:")
+            logger.error(f"  Error type: {type(e).__name__}")
+            logger.error(f"  Error message: {str(e)}")
+            logger.error(f"  Batch keys: {list(batch.keys())}")
+            if audio_column_name in batch:
+                audio_data = batch[audio_column_name]
+                logger.error(f"  Audio type: {type(audio_data)}")
+                if isinstance(audio_data, dict):
+                    logger.error(f"  Audio keys: {list(audio_data.keys())}")
+                elif hasattr(audio_data, '__dict__'):
+                    logger.error(f"  Audio attributes: {dir(audio_data)}")
+            raise
 
     with training_args.main_process_first(desc="dataset map pre-processing"):
         vectorized_datasets = raw_datasets.map(
@@ -490,6 +563,16 @@ def main():
             remove_columns=raw_datasets_features,
         ).with_format("torch")
 
+        # Inspect vectorized dataset
+        for split, dataset in vectorized_datasets.items():
+            try:
+                first_example = next(iter(dataset))
+                logger.info(f"First example from {split} after processing:")
+                logger.info(f"Keys: {first_example.keys()}")
+                logger.info(f"Types: {', '.join(f'{k}: {type(v)}' for k, v in first_example.items())}")
+            except Exception as e:
+                logger.warning(f"Could not inspect first example from vectorized {split}: {e}")
+
         if training_args.do_train and data_args.streaming:
             # manually shuffle if streaming (done by the trainer for non-streaming)
             vectorized_datasets["train"] = vectorized_datasets["train"].shuffle(
@@ -551,7 +634,13 @@ def main():
     # Trainer callback to reinitialise and reshuffle the streamable datasets at the beginning of each epoch
     # Only required for streaming: Trainer automatically shuffles non-streaming datasets
     class ShuffleCallback(TrainerCallback):
-        def on_epoch_begin(self, args, state, control, train_dataloader, **kwargs):
+        def on_train_begin(self, args, state, control, **kwargs):
+            self.trainer = kwargs.get('trainer')
+
+        def on_epoch_begin(self, args, state, control, **kwargs):
+            if not hasattr(self, "trainer") or not hasattr(self.trainer, "train_dataloader") or self.trainer.train_dataloader is None:
+                return
+            train_dataloader = self.trainer.train_dataloader
             if isinstance(train_dataloader.dataset, IterableDatasetShard):
                 pass  # set_epoch() is handled by the Trainer
             elif isinstance(train_dataloader.dataset, IterableDataset):

run_speech_recognition_seq2seq_streaming_cv.py

@@ -0,0 +1,657 @@
+#!/usr/bin/env python
+# coding=utf-8
+# Copyright 2022 The HuggingFace Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""
+Fine-tuning the library models for sequence to sequence speech recognition
+with 🤗 Datasets' streaming mode.
+"""
+# You can also adapt this script for your own sequence to sequence speech
+# recognition task. Pointers for this are left as comments.
+
+import logging
+import os
+import sys
+from dataclasses import dataclass, field
+from typing import Any, Dict, List, Optional, Union
+import numpy
+
+import datasets
+import torch
+from datasets import DatasetDict, IterableDatasetDict, interleave_datasets, load_dataset
+from torch.utils.data import IterableDataset
+
+import evaluate
+import transformers
+from transformers import (
+    AutoConfig,
+    AutoFeatureExtractor,
+    AutoModelForSpeechSeq2Seq,
+    AutoProcessor,
+    AutoTokenizer,
+    HfArgumentParser,
+    Seq2SeqTrainer,
+    Seq2SeqTrainingArguments,
+    TrainerCallback,
+    set_seed,
+)
+from transformers.models.whisper.english_normalizer import BasicTextNormalizer
+from transformers.trainer_pt_utils import IterableDatasetShard
+from transformers.trainer_utils import get_last_checkpoint, is_main_process
+from transformers.utils import check_min_version, send_example_telemetry
+from transformers.utils.versions import require_version
+
+
+# Will error if the minimal version of Transformers is not installed. Remove at your own risks.
+check_min_version("4.25.0.dev0")
+
+require_version("datasets>=1.18.2", "To fix: pip install -r examples/pytorch/speech-recognition/requirements.txt")
+
+logger = logging.getLogger(__name__)
+
+
+@dataclass
+class ModelArguments:
+    """
+    Arguments pertaining to which model/config/tokenizer we are going to fine-tune from.
+    """
+
+    model_name_or_path: str = field(
+        metadata={"help": "Path to pretrained model or model identifier from huggingface.co/models"}
+    )
+    config_name: Optional[str] = field(
+        default=None, metadata={"help": "Pretrained config name or path if not the same as model_name"}
+    )
+    tokenizer_name: Optional[str] = field(
+        default=None, metadata={"help": "Pretrained tokenizer name or path if not the same as model_name"}
+    )
+    feature_extractor_name: Optional[str] = field(
+        default=None, metadata={"help": "feature extractor name or path if not the same as model_name"}
+    )
+    cache_dir: Optional[str] = field(
+        default=None,
+        metadata={"help": "Where to store the pretrained models downloaded from huggingface.co"},
+    )
+    use_fast_tokenizer: bool = field(
+        default=True,
+        metadata={"help": "Whether to use one of the fast tokenizer (backed by the tokenizers library) or not."},
+    )
+    model_revision: str = field(
+        default="main",
+        metadata={"help": "The specific model version to use (can be a branch name, tag name or commit id)."},
+    )
+    use_auth_token: bool = field(
+        default=False,
+        metadata={
+            "help": (
+                "Will use the token generated when running `huggingface-cli login` (necessary to use this script "
+                "with private models)."
+            )
+        },
+    )
+    freeze_feature_encoder: bool = field(
+        default=True, metadata={"help": "Whether to freeze the feature encoder layers of the model."}
+    )
+    freeze_encoder: bool = field(
+        default=False, metadata={"help": "Whether to freeze the entire encoder of the seq2seq model."}
+    )
+    forced_decoder_ids: List[List[int]] = field(
+        default=None,
+        metadata={
+            "help": (
+                "A list of pairs of integers which indicates a mapping from generation indices to token indices "
+                "that will be forced before sampling. For example, [[0, 123]] means the first generated token "
+                "will always be a token of index 123."
+            )
+        },
+    )
+    suppress_tokens: List[int] = field(
+        default=None, metadata={"help": "A list of tokens that will be suppressed at generation."}
+    )
+    model_index_name: str = field(default=None, metadata={"help": "Pretty name for the model card."})
+
+
+@dataclass
+class DataTrainingArguments:
+    """
+    Arguments pertaining to what data we are going to input our model for training and eval.
+    """
+
+    dataset_name: str = field(
+        default=None, metadata={"help": "The name of the dataset to use (via the datasets library)."}
+    )
+    dataset_config_name: Optional[str] = field(
+        default=None, metadata={"help": "The configuration name of the dataset to use (via the datasets library)."}
+    )
+    text_column: Optional[str] = field(
+        default=None,
+        metadata={"help": "The name of the column in the datasets containing the full texts (for summarization)."},
+    )
+    max_train_samples: Optional[int] = field(
+        default=None,
+        metadata={
+            "help": (
+                "For debugging purposes or quicker training, truncate the number of training examples to this "
+                "value if set."
+            )
+        },
+    )
+    max_eval_samples: Optional[int] = field(
+        default=None,
+        metadata={
+            "help": (
+                "For debugging purposes or quicker training, truncate the number of evaluation examples to this "
+                "value if set."
+            )
+        },
+    )
+    audio_column_name: str = field(
+        default="audio",
+        metadata={"help": "The name of the dataset column containing the audio data. Defaults to 'audio'"},
+    )
+    text_column_name: str = field(
+        default="text",
+        metadata={"help": "The name of the dataset column containing the text data. Defaults to 'text'"},
+    )
+    max_duration_in_seconds: float = field(
+        default=20.0,
+        metadata={
+            "help": (
+                "Truncate audio files that are longer than `max_duration_in_seconds` seconds to"
+                " 'max_duration_in_seconds`"
+            )
+        },
+    )
+    min_duration_in_seconds: float = field(
+        default=0.0, metadata={"help": "Filter audio files that are shorter than `min_duration_in_seconds` seconds"}
+    )
+    train_split_name: str = field(
+        default="train",
+        metadata={
+            "help": "The name of the training data set split to use (via the datasets library). Defaults to 'train'"
+        },
+    )
+    eval_split_name: str = field(
+        default="test",
+        metadata={
+            "help": "The name of the training data set split to use (via the datasets library). Defaults to 'train'"
+        },
+    )
+    do_lower_case: bool = field(
+        default=False,
+        metadata={"help": "Whether the target text should be lower cased."},
+    )
+    do_remove_punctuation: bool = field(
+        default=False,
+        metadata={"help": "Whether the target text should be striped of punctuation."},
+    )
+    do_normalize_eval: bool = field(
+        default=True,
+        metadata={"help": "Whether to normalise the references and predictions in the eval WER calculation."},
+    )
+    language: str = field(
+        default=None,
+        metadata={
+            "help": (
+                "Language for multilingual fine-tuning. This argument should be set for multilingual fine-tuning "
+                "only. For English speech recognition, it should be set to `None`."
+            )
+        },
+    )
+    task: str = field(
+        default="transcribe",
+        metadata={"help": "Task, either `transcribe` for speech recognition or `translate` for speech translation."},
+    )
+    shuffle_buffer_size: Optional[int] = field(
+        default=500,
+        metadata={
+            "help": (
+                "The number of streamed examples to download before shuffling them. The large the buffer, "
+                "the closer it is to real offline shuffling."
+            )
+        },
+    )
+    streaming: bool = field(
+        default=True,
+        metadata={"help": "Whether to use streaming mode to load and pre-process the data."},
+    )
+
+
+@dataclass
+class DataCollatorSpeechSeq2SeqWithPadding:
+    """
+    Data collator that will dynamically pad the inputs received.
+    Args:
+        processor ([`WhisperProcessor`])
+            The processor used for processing the data.
+        decoder_start_token_id (`int`)
+            The begin-of-sentence of the decoder.
+    """
+
+    processor: Any
+    decoder_start_token_id: int
+
+    def __call__(self, features: List[Dict[str, Union[List[int], torch.Tensor]]]) -> Dict[str, torch.Tensor]:
+        # split inputs and labels since they have to be of different lengths and need
+        # different padding methods
+        model_input_name = self.processor.model_input_names[0]
+        input_features = [{model_input_name: feature[model_input_name]} for feature in features]
+        label_features = [{"input_ids": feature["labels"]} for feature in features]
+
+        batch = self.processor.feature_extractor.pad(input_features, return_tensors="pt")
+
+        labels_batch = self.processor.tokenizer.pad(label_features, return_tensors="pt")
+
+        # replace padding with -100 to ignore loss correctly
+        labels = labels_batch["input_ids"].masked_fill(labels_batch.attention_mask.ne(1), -100)
+
+        # if bos token is appended in previous tokenization step,
+        # cut bos token here as it's append later anyways
+        if (labels[:, 0] == self.decoder_start_token_id).all().cpu().item():
+            labels = labels[:, 1:]
+
+        batch["labels"] = labels
+
+        return batch
+
+
+def load_maybe_streaming_dataset(dataset_name, dataset_config_name, split="train", streaming=True, **kwargs):
+    """
+    Utility function to load a dataset in streaming mode. For datasets with multiple splits,
+    each split is loaded individually and then splits combined by taking alternating examples from
+    each (interleaving).
+    """
+    if ("+" in split):
+        # load multiple splits separated by the `+` symbol with streaming mode
+        dataset_splits = [
+            load_dataset(dataset_name, dataset_config_name, split=split_name, streaming=streaming, trust_remote_code=True, **kwargs)
+            for split_name in split.split("+")
+        ]
+        # interleave multiple splits to form one dataset
+        interleaved_dataset = interleave_datasets(dataset_splits)
+        return interleaved_dataset
+    else:
+        # load a single split *with* streaming mode
+        dataset = load_dataset(dataset_name, dataset_config_name, split=split, streaming=streaming, trust_remote_code=True, **kwargs)
+        return dataset
+
+
+def main():
+    # 1. Parse input arguments
+    # See all possible arguments in src/transformers/training_args.py
+    # or by passing the --help flag to this script.
+    # We now keep distinct sets of args, for a cleaner separation of concerns.
+    parser = HfArgumentParser((ModelArguments, DataTrainingArguments, Seq2SeqTrainingArguments))
+
+    if len(sys.argv) == 2 and sys.argv[1].endswith(".json"):
+        # If we pass only one argument to the script and it's the path to a json file,
+        # let's parse it to get our arguments.
+        model_args, data_args, training_args = parser.parse_json_file(json_file=os.path.abspath(sys.argv[1]))
+    else:
+        model_args, data_args, training_args = parser.parse_args_into_dataclasses()
+
+    # Sending telemetry. Tracking the example usage helps us better allocate resources to maintain them. The
+    # information sent is the one passed as arguments along with your Python/PyTorch versions.
+    send_example_telemetry("run_speech_recognition_seq2seq_streaming", model_args, data_args)
+
+    # 2. Setup logging
+    logging.basicConfig(
+        format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
+        datefmt="%m/%d/%Y %H:%M:%S",
+        handlers=[logging.StreamHandler(sys.stdout)],
+    )
+    log_level = training_args.get_process_log_level()
+    logger.setLevel(log_level)
+    datasets.utils.logging.set_verbosity(log_level)
+    transformers.utils.logging.set_verbosity(log_level)
+    transformers.utils.logging.enable_default_handler()
+    transformers.utils.logging.enable_explicit_format()
+
+    logger.setLevel(logging.INFO if is_main_process(training_args.local_rank) else logging.WARN)
+
+    # Log on each process the small summary:
+    logger.warning(
+        f"Process rank: {training_args.local_rank}, device: {training_args.device}, n_gpu: {training_args.n_gpu}"
+        f"distributed training: {bool(training_args.local_rank != -1)}, 16-bits training: {training_args.fp16}"
+    )
+    logger.info(f"Training/evaluation parameters {training_args}")
+
+    # Set the verbosity to info of the Transformers logger (on main process only):
+    if is_main_process(training_args.local_rank):
+        transformers.utils.logging.set_verbosity_info()
+    logger.info("Training/evaluation parameters %s", training_args)
+
+    # 3. Detecting last checkpoint and eventually continue from last checkpoint
+    last_checkpoint = None
+    if os.path.isdir(training_args.output_dir) and training_args.do_train and not training_args.overwrite_output_dir:
+        last_checkpoint = get_last_checkpoint(training_args.output_dir)
+        if last_checkpoint is None and len(os.listdir(training_args.output_dir)) > 0:
+            raise ValueError(
+                f"Output directory ({training_args.output_dir}) already exists and is not empty. "
+                "Use --overwrite_output_dir to overcome."
+            )
+        elif last_checkpoint is not None and training_args.resume_from_checkpoint is None:
+            logger.info(
+                f"Checkpoint detected, resuming training at {last_checkpoint}. To avoid this behavior, change "
+                "the `--output_dir` or add `--overwrite_output_dir` to train from scratch."
+            )
+
+    # Set seed before initializing model.
+    set_seed(training_args.seed)
+
+    # 4. Load dataset
+    raw_datasets = IterableDatasetDict() if data_args.streaming else DatasetDict()
+
+    if training_args.do_train:
+        raw_datasets["train"] = load_maybe_streaming_dataset(
+            data_args.dataset_name,
+            data_args.dataset_config_name,
+            split=data_args.train_split_name,
+            #use_auth_token=True if model_args.use_auth_token else None,
+            streaming=data_args.streaming,
+        )
+
+    if training_args.do_eval:
+        raw_datasets["eval"] = load_maybe_streaming_dataset(
+            data_args.dataset_name,
+            data_args.dataset_config_name,
+            split=data_args.eval_split_name,
+            #use_auth_token=True if model_args.use_auth_token else None,
+            streaming=data_args.streaming,
+        )
+
+    raw_datasets_features = list(next(iter(raw_datasets.values())).features.keys())
+
+    if data_args.audio_column_name not in raw_datasets_features:
+        raise ValueError(
+            f"--audio_column_name '{data_args.audio_column_name}' not found in dataset '{data_args.dataset_name}'. "
+            "Make sure to set `--audio_column_name` to the correct audio column - one of "
+            f"{', '.join(raw_datasets_features)}."
+        )
+
+    if data_args.text_column_name not in raw_datasets_features:
+        raise ValueError(
+            f"--text_column_name {data_args.text_column_name} not found in dataset '{data_args.dataset_name}'. "
+            "Make sure to set `--text_column_name` to the correct text column - one of "
+            f"{', '.join(raw_datasets_features)}."
+        )
+
+    # 5. Load pretrained model, tokenizer, and feature extractor
+    #
+    # Distributed training:
+    # The .from_pretrained methods guarantee that only one local process can concurrently
+    config = AutoConfig.from_pretrained(
+        model_args.config_name if model_args.config_name else model_args.model_name_or_path,
+        cache_dir=model_args.cache_dir,
+        revision=model_args.model_revision,
+        use_auth_token=True if model_args.use_auth_token else None,
+    )
+
+    config.update({"forced_decoder_ids": model_args.forced_decoder_ids, "suppress_tokens": model_args.suppress_tokens})
+
+    if training_args.gradient_checkpointing:
+        config.update({"use_cache": False})
+
+    feature_extractor = AutoFeatureExtractor.from_pretrained(
+        model_args.feature_extractor_name if model_args.feature_extractor_name else model_args.model_name_or_path,
+        cache_dir=model_args.cache_dir,
+        revision=model_args.model_revision,
+        use_auth_token=True if model_args.use_auth_token else None,
+    )
+    tokenizer = AutoTokenizer.from_pretrained(
+        model_args.tokenizer_name if model_args.tokenizer_name else model_args.model_name_or_path,
+        cache_dir=model_args.cache_dir,
+        use_fast=model_args.use_fast_tokenizer,
+        revision=model_args.model_revision,
+        use_auth_token=True if model_args.use_auth_token else None,
+    )
+    model = AutoModelForSpeechSeq2Seq.from_pretrained(
+        model_args.model_name_or_path,
+        config=config,
+        cache_dir=model_args.cache_dir,
+        revision=model_args.model_revision,
+        use_auth_token=True if model_args.use_auth_token else None,
+    )
+
+    if model.config.decoder_start_token_id is None:
+        raise ValueError("Make sure that `config.decoder_start_token_id` is correctly defined")
+
+    if model_args.freeze_feature_encoder:
+        model.freeze_feature_encoder()
+
+    if model_args.freeze_encoder:
+        model.freeze_encoder()
+
+    if data_args.language is not None:
+        # We only need to set the task id when the language is specified (i.e. in a multilingual setting)
+        tokenizer.set_prefix_tokens(language=data_args.language, task=data_args.task)
+
+    # 6. Resample speech dataset if necessary
+    dataset_sampling_rate = next(iter(raw_datasets.values())).features[data_args.audio_column_name].sampling_rate
+    if dataset_sampling_rate != feature_extractor.sampling_rate:
+        raw_datasets = raw_datasets.cast_column(
+            data_args.audio_column_name, datasets.features.Audio(sampling_rate=feature_extractor.sampling_rate)
+        )
+
+    # 7. Preprocessing the datasets.
+    # We need to read the audio files as arrays and tokenize the targets.
+    max_input_length = data_args.max_duration_in_seconds * feature_extractor.sampling_rate
+    min_input_length = data_args.min_duration_in_seconds * feature_extractor.sampling_rate
+    audio_column_name = data_args.audio_column_name
+    text_column_name = data_args.text_column_name
+    model_input_name = feature_extractor.model_input_names[0]
+    do_lower_case = data_args.do_lower_case
+    do_remove_punctuation = data_args.do_remove_punctuation
+    normalizer = BasicTextNormalizer()  # 'official' text normalizer from OpenAI
+
+    if data_args.max_train_samples is not None:
+        raw_datasets["train"] = (
+            raw_datasets["train"].take(data_args.max_train_samples)
+            if data_args.streaming
+            else raw_datasets["train"].select(range(data_args.max_train_samples))
+        )
+
+    if data_args.max_eval_samples is not None:
+        raw_datasets["eval"] = (
+            raw_datasets["eval"].take(data_args.max_eval_samples)
+            if data_args.streaming
+            else raw_datasets["eval"].select(range(data_args.max_eval_samples))
+        )
+
+    def prepare_dataset(batch):
+        # process audio
+        sample = batch[audio_column_name]
+        
+        # Handle different audio formats - some datasets provide raw arrays, others provide paths
+        if isinstance(sample, dict):
+            if "array" in sample:
+                audio_array = sample["array"]
+                sampling_rate = sample["sampling_rate"]
+            elif "path" in sample:
+                # Load from path if array is not available
+                audio_array = sample["path"]  # datasets will load the file for us
+                sampling_rate = sample.get("sampling_rate", feature_extractor.sampling_rate)
+            else:
+                raise ValueError(f"Unsupported audio format. Sample must contain either 'array' or 'path'. Got {sample.keys()}")
+        else:
+            # Assume it's a direct path or array
+            audio_array = sample
+            sampling_rate = feature_extractor.sampling_rate
+
+        inputs = feature_extractor(audio_array, sampling_rate=sampling_rate)
+        
+        # process audio length
+        if isinstance(audio_array, numpy.ndarray):
+            batch["input_length"] = len(audio_array)
+        else:
+            # If we couldn't get the direct array length, estimate it from the processed features
+            batch["input_length"] = inputs.get(model_input_name)[0].shape[0] * feature_extractor.hop_length
+
+        # process targets
+        input_str = batch[text_column_name].lower() if do_lower_case else batch[text_column_name]
+        if do_remove_punctuation:
+            input_str = normalizer(input_str).strip()
+        batch["labels"] = tokenizer(input_str).input_ids
+        return batch
+
+    with training_args.main_process_first(desc="dataset map pre-processing"):
+        vectorized_datasets = raw_datasets.map(
+            prepare_dataset,
+            remove_columns=raw_datasets_features,
+        ).with_format("torch")
+
+        if training_args.do_train and data_args.streaming:
+            # manually shuffle if streaming (done by the trainer for non-streaming)
+            vectorized_datasets["train"] = vectorized_datasets["train"].shuffle(
+                buffer_size=data_args.shuffle_buffer_size,
+                seed=training_args.seed,
+            )
+
+    # filter training data that is shorter than min_input_length or longer than
+    # max_input_length
+    def is_audio_in_length_range(length):
+        return min_input_length < length < max_input_length
+
+    if training_args.do_train:
+        vectorized_datasets["train"] = vectorized_datasets["train"].filter(
+            is_audio_in_length_range,
+            input_columns=["input_length"],
+        )
+
+    # 8. Load Metric
+    metric = evaluate.load("wer")
+    do_normalize_eval = data_args.do_normalize_eval
+
+    def compute_metrics(pred):
+        pred_ids = pred.predictions
+
+        pred.label_ids[pred.label_ids == -100] = tokenizer.pad_token_id
+
+        pred_str = tokenizer.batch_decode(pred_ids, skip_special_tokens=True)
+        # we do not want to group tokens when computing the metrics
+        label_str = tokenizer.batch_decode(pred.label_ids, skip_special_tokens=True)
+
+        if do_normalize_eval:
+            pred_str = [normalizer(pred) for pred in pred_str]
+            label_str = [normalizer(label) for label in label_str]
+            # filtering step to only evaluate the samples that correspond to non-zero references:
+            pred_str = [pred_str[i] for i in range(len(pred_str)) if len(label_str[i]) > 0]
+            label_str = [label_str[i] for i in range(len(label_str)) if len(label_str[i]) > 0]
+
+        wer = 100 * metric.compute(predictions=pred_str, references=label_str)
+
+        return {"wer": wer}
+
+    # 9. Create a single speech processor
+    if is_main_process(training_args.local_rank):
+        # save feature extractor, tokenizer and config
+        feature_extractor.save_pretrained(training_args.output_dir)
+        tokenizer.save_pretrained(training_args.output_dir)
+        config.save_pretrained(training_args.output_dir)
+
+    processor = AutoProcessor.from_pretrained(training_args.output_dir)
+
+    # 10. Define data collator
+    data_collator = DataCollatorSpeechSeq2SeqWithPadding(
+        processor=processor,
+        decoder_start_token_id=model.config.decoder_start_token_id,
+    )
+
+    # 11. Configure Trainer
+    # Trainer callback to reinitialise and reshuffle the streamable datasets at the beginning of each epoch
+    # Only required for streaming: Trainer automatically shuffles non-streaming datasets
+    class ShuffleCallback(TrainerCallback):
+        def on_train_begin(self, args, state, control, **kwargs):
+            self.trainer = kwargs.get('trainer')
+
+        def on_epoch_begin(self, args, state, control, **kwargs):
+            if not hasattr(self, "trainer") or not hasattr(self.trainer, "train_dataloader") or self.trainer.train_dataloader is None:
+                return
+            train_dataloader = self.trainer.train_dataloader
+            if isinstance(train_dataloader.dataset, IterableDatasetShard):
+                pass  # set_epoch() is handled by the Trainer
+            elif isinstance(train_dataloader.dataset, IterableDataset):
+                train_dataloader.dataset.set_epoch(train_dataloader.dataset._epoch + 1)
+
+    # Initialize Trainer
+    trainer = Seq2SeqTrainer(
+        model=model,
+        args=training_args,
+        train_dataset=vectorized_datasets["train"] if training_args.do_train else None,
+        eval_dataset=vectorized_datasets["eval"] if training_args.do_eval else None,
+        processing_class=feature_extractor,
+        data_collator=data_collator,
+        compute_metrics=compute_metrics if training_args.predict_with_generate else None,
+        callbacks=[ShuffleCallback()] if data_args.streaming else None,
+    )
+
+    # 12. Training
+    if training_args.do_train:
+        checkpoint = None
+        if training_args.resume_from_checkpoint is not None:
+            checkpoint = training_args.resume_from_checkpoint
+        elif last_checkpoint is not None:
+            checkpoint = last_checkpoint
+        train_result = trainer.train(resume_from_checkpoint=checkpoint)
+        trainer.save_model()  # Saves the feature extractor too for easy upload
+
+        metrics = train_result.metrics
+        if data_args.max_train_samples:
+            metrics["train_samples"] = data_args.max_train_samples
+        trainer.log_metrics("train", metrics)
+        trainer.save_metrics("train", metrics)
+        trainer.save_state()
+
+    # 13. Evaluation
+    results = {}
+    if training_args.do_eval:
+        logger.info("*** Evaluate ***")
+        metrics = trainer.evaluate(
+            metric_key_prefix="eval",
+            max_length=training_args.generation_max_length,
+            num_beams=training_args.generation_num_beams,
+        )
+        if data_args.max_eval_samples:
+            metrics["eval_samples"] = data_args.max_eval_samples
+
+        trainer.log_metrics("eval", metrics)
+        trainer.save_metrics("eval", metrics)
+
+    # 14. Write Training Stats
+    kwargs = {
+        "finetuned_from": model_args.model_name_or_path,
+        "tasks": "automatic-speech-recognition",
+        "tags": "whisper-event",
+    }
+    if data_args.dataset_name is not None:
+        kwargs["dataset_tags"] = data_args.dataset_name
+        if data_args.dataset_config_name is not None:
+            kwargs["dataset"] = f"{data_args.dataset_name} {data_args.dataset_config_name}"
+        else:
+            kwargs["dataset"] = data_args.dataset_name
+        if "common_voice" in data_args.dataset_name:
+            kwargs["language"] = data_args.dataset_config_name.split('-')[0]
+        if model_args.model_index_name is not None:
+            kwargs["model_name"] = model_args.model_index_name
+
+    if training_args.push_to_hub:
+        trainer.push_to_hub(**kwargs)
+    else:
+        trainer.create_model_card(**kwargs)
+
+    return results
+
+
+if __name__ == "__main__":
+    main()

tokenizer_config.json

@@ -12987,6 +12987,7 @@
   "clean_up_tokenization_spaces": true,
   "eos_token": "<|endoftext|>",
   "errors": "replace",
+  "extra_special_tokens": {},
   "model_max_length": 1000000000000000019884624838656,
   "pad_token": "<|endoftext|>",
   "processor_class": "WhisperProcessor",

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:bf7a986abd3107068a3a0d5d4f994d9792d101aaf76cd5472057cde7be944e18
-size 5368
+oid sha256:e1fbb1e83bfa724f802875b50dc42275d9c7a8a5ff867993f1edbeedd5b91d4e
+size 5432